CSE 401 Wi13 - MiniJava

Overview

The course project is to implement the MiniJava language specified in the Appendix of Appel's Modern Compiler Implementation in Java, 2nd edition and described on the MiniJava web site. MiniJava is a subset of Java and the meaning of a MiniJava program is given by its meaning as a Java program.

Compared to full Java, a MiniJava program consists of a single class containing a static main method followed by zero or more other classes. There are no other static methods or variables. Classes may extend other classes, and method overriding is included, but method overloading is not. All classes making up a MiniJava program are included in a single source file. The only types available are int, boolean, int[] and reference types (classes). "System.out.println(...);" is a statement and can only print integers - it is not a normal method call, and does not refer to a static method of a class. All methods are value-returning and must end with a single return statement. The only other available statements are if, while, and assignment. There are other simplifications to keep the project size reasonable.

The full MiniJava grammar is given on the project web site and in the Appendix of Appel's Modern Compiler Implementation in Java (2nd ed). Look at the grammar carefully to see what is and is not included in the language subset. You should implement full MiniJava as described there, except that you do not need to implement nested /* comments */. (You need to implement /* */ comments, but do not need to allow them to nest, i.e., the first */ terminates any open comment regardless of how many /* symbols have appeared before it, as in standard Java. You also need to implement // comments.)

There are two symbols in the grammar that are not otherwise specified. An <IDENTIFIER> is a sequence of letters, digits, and underscores, starting with a letter. Uppercase letters are distinguished from lowercase. An <INTEGER_LITERAL> is a sequence of decimal digits that denotes an integer value.

Implementation

Your MiniJava compiler should be implemented in Java using the JFlex/CUP scanner/parser tools. You are free to use any development platform and environment you wish, but your resulting project should build (using ant) and run on the lab linux machines, attu, or the lab VM (more details forthcoming). If you use Eclipse, you can create a suitable project by following the instructions in the starter code README files. Don't create a new generic Java project - it doesn't set things up correctly.

Extensions

The basic project requirements are small enough to be tractable in a one-quarter course project, but include enough to cover the core ideas of compiling an object-oriented (as well as procedural) language. If you are feeling ambitous and have the time, you are invited to add additional parts of Java to the language. Here are a few suggestions.

Some Simple Ideas

• Add nested /* ... */ comments.
• Add additional arithmetic and relational operators for integers that are not included in the MiniJava specification.
• Add null as a constant expression and support == and != for object reference types.
• Allow return statements anywhere in a method.
• Allow calls of local methods without the explicit use of this.
• Relax the ordering of declarations so that variables can be declared anywhere in a method body.
• Allow initialization of variables in declarations.
• Add void methods, a return statement with no expression, and appropriate type checking rules.
• Support public and private declarations on both methods and instance variables, and check to ensure that access restrictions are not violated.

Adding an additional numeric type like double would not be particularly interesting, since there is relatively little difference in how the compiler handles ints and doubles. However, adding String literals and the concatenation (+) operator would be interesting.

More Sophisticated, but very interesting

• Support instanceof and type casts (which can require a runtime instanceof check).
• Support super. as a prefix in method calls.
• Add constructor declarations with optional parameters. To make this useful, you probably want to include the use of super(...) at the beginning of the body of a constructor.
• Support arrays of objects (no harder to implement than arrays of ints, except the type checking rules are more interesting).

Of the suggested extensions, adding instanceof, type casts, and super. are particularly instructive.

Some small amount of extra credit will be awarded to projects that go beyond the basic requirements.